Optimising Runtime Safety Analysis Efficiency for Self-Organising Systems

Abstract

Self-organising resource-flow systems typically havea high tolerance for component faults: When a component fails, the system can use another component of the same type instead. However, this redundancy is eventually exhausted: If enoughcomponents fail, they can no longer be replaced and the systemceases to function. An analysis of these self-organisation limitsis essential to assess the system's safety but difficult to performat design time because the system's structure and behaviour arehard to predict. By contrast, runtime analyses are subject to highperformance demands. This paper presents several techniquesthat significantly reduce analysis time in order to facilitate safetyanalyses at runtime. We model a self-organising system producingpersonalised medicine and use it to evaluate these techniques.